Show HN: TraceML, a tool to trace live memory usage in PyTorch training

1 comments

• $traceopt-ai 10 hours ago

  I built this after wasting hours on OOM errors, curious if others would want optimizer/efficiency hints next?